The future behaviour of the global ocean as a sink for CO2 is significant for climate change, but it is also important to understand its past by quantifying anthropogenic CO2 (Cant) in the ocean today. Unfortunately, this is complicated by the difficulty of deconvoluting Cant from the natural, unperturbed carbon cycle. Nonetheless, a range of techniques have been devised that perform this separation using the information implicit in other physical, biogeochemical and artificial ocean tracers. One such method is the TrOCA approach, whose parameterisation is derived from relationships between biogeochemical tracers within watermasses defined by age tracers such as CFC-11. TrOCA has a number of methodological advantages, and has been shown to be plausible, relative to other methods, in a number of studies. Here we examine the TrOCA approach by using it to deconvolute the known distribution of Cant from an ocean general circulation model (OGCM) simulation of the industrial period (1864–2004). TrOCA is evaluated at local, regional and global scales, with an emphasis on the wider applicability of the parameterisations derived at these scales. Our work finds that the published TrOCA parameterisation performs poorly when extrapolated beyond its calibration region, either with observational data or (especially) model output. Optimising TrOCA parameters using model output as a synthetic dataset leads to some small improvements, but the resulting TrOCA variants still perform poorly. Furthermore, there are large ranges on the optimised TrOCA parameters suggesting that a "universal" TrOCA parameterisation is not achieveable.